Perception-Aware Risk-Constrained Replication and Roaming for Reliable Wireless TSN Flows in Mobile Robotics

Abstract

We present PARR, a perception-aware, risk-constrained replication-and-roaming policy for wireless links carrying time-sensitive-networking (TSN) flows in mobile robotics. PARR is not a MAC scheduler or a standalone predictor: IEEE 802.1CB frame replication (FRER) provides reliability, while LiDAR geometry and an online-calibrated shadow-fading residual decide when, where, and over which access points (APs) to replicate at minimum airtime. On a 3GPP TR 38.901 indoor-factory channel with correlated shadowing, PARR matches blanket-FRER reliability at about one-third of its airtime, stays robust as shadowing grows where threshold and reactive baselines fail, and, unlike blanket FRER which is harmful under contention, scales to sixteen coexisting robots.